Reference Information › CA AppLogic Support Knowledge Base › Overview of Support Knowledge Base › Tuning the Controller

Tuning the Controller

The AppLogic controller is a bottleneck in the system. It supports all grid level activity, the webUI, editing operations like appliance design and building, and when you have multiple users attached to the controller, overall grid performance can suffer at times.

There are adjustments you can make to tune the controller, and practices you can implement as discussed below:

Grid Parameters

A few grid parameters are available to add resources to the controller.

• ctl_blk_copy – Volume copies for block-level volumes are performed on the grid controller by default; if this parameter is set to 0, the volume copies are offloaded to a filer, thus relieving the grid controller of the task. Use this parameter when there will be multiple users simultaneously logged into the grid controller and performing operations that require block-level volume copies (e.g. provisioning and copying applications without volume resize).
• ctl_vm_mb – The amount of memory available to the controller. This can be increased to devote more server memory to the controller, which normally uses 768 Megabytes.
• ctl_vm_cpu - The percentage of the server CPU time allocated for the controller. This is the maximum that the controller can take, in case the CPU is scheduled to run appliance VMs. The server information for the host on which the controller is running will reflect this by showing less than 100% available resource for that server.

Primary/Secondary Selection

The primary designated server is where the controller runs; secondary servers are where it goes if something happens to the primary. If you name an under-resourced server as primary, it will run poorly; if you make it secondary, it may end up running the controller and the controller will run poorly. Make sure your primary and secondary servers are equivalent and configure resource-deficient servers with no primary or secondary role.

Disk Striping

When using multiple large disks on the grid servers, striping those disks during the grid propagation can make disk response a bit faster, increasing overall performance of the grid.